Technique-to-Study Mapping

This page provides the complete study-level mapping for the normalized technique families used in RQ2 of the survey Fuzzing AI Systems: Foundations, Techniques, and Open Challenges.

The analysis includes 125 primary studies identified within the January 2015–February 2026 search window. Technique-family labels are not mutually exclusive because many studies combine multiple central fuzzing mechanisms.

Each paper ID is linked to its corresponding entry on the Primary Studies page, where the full title, authors, year, venue, and DOI or publisher URL are provided.

Detailed Mapping

Normalized technique family	# Studies	Primary-study IDs
Mutation-based	63	P004, P005, P006, P007, P008, P009, P010, P011, P014, P016, P018, P021, P024, P026, P031, P032, P034, P035, P038, P043, P044, P051, P058, P060, P061, P066, P068, P069, P076, P088, P089, P103, P106, P116, P125, P130, P133, P136, P137, P144, P145, P146, P148, P152, P154, P160, P164, P167, P169, P170, P172, P174, P176, P178, P180, P182, P193, P197, P198, P200, P201, P204, P208
Coverage-guided	52	P004, P007, P012, P019, P024, P025, P027, P028, P029, P031, P032, P033, P039, P040, P042, P045, P046, P061, P066, P080, P086, P096, P104, P120, P121, P125, P130, P133, P140, P142, P143, P144, P145, P146, P148, P153, P154, P157, P158, P162, P165, P167, P174, P176, P178, P180, P182, P197, P200, P201, P202, P204
Learning-based	30	P002, P012, P013, P015, P023, P026, P030, P035, P068, P078, P079, P081, P082, P090, P099, P101, P103, P104, P114, P122, P125, P127, P130, P133, P136, P143, P145, P146, P147, P152
Constraint-guided	17	P001, P003, P005, P006, P008, P011, P017, P020, P076, P077, P100, P111, P113, P114, P147, P172, P198
Search/Heuristic-guided	16	P018, P020, P022, P029, P030, P035, P038, P044, P045, P058, P069, P083, P087, P145, P152, P193
Differential testing	15	P001, P003, P008, P009, P010, P015, P016, P018, P091, P095, P119, P156, P170, P193, P204
LLM/Prompt-guided	14	P002, P013, P015, P023, P078, P081, P082, P090, P099, P101, P103, P104, P122, P127
Generation-based	7	P025, P028, P034, P036, P119, P143, P147
Metamorphic testing	3	P039, P041, P049
Debugging/Fault-localization	1	P050

Note: Technique-family labels are not mutually exclusive because some studies combine multiple central mechanisms. This table uses the normalized high-level labels used in RQ2. Original descriptive labels and normalization rationales are preserved in the annotated dataset available through the replication package.

Interpretation

The mapping shows that mutation-based and coverage-guided techniques form the largest and most widely combined families. Learning-based, constraint-guided, search/heuristic-guided, differential, and LLM/prompt-guided mechanisms often appear as complementary forms of generation, guidance, or failure detection.

The overlap across rows reflects the hybrid nature of AI-system fuzzing. A single study may, for example, combine mutation with coverage feedback, constraint-guided generation with differential testing, or LLM-based generation with validation and comparison mechanisms.

The complete machine-readable annotations are available in the replication package on Zenodo.